Follow

people in work debating the ethics of automated cars and if they have to make a trolley problem like decision and god people have so much faith in AI to actually solving the problem given. we're not going to have cars that decide to hit a pensioner over toddlers we're going to have cars that decide people in green cardigans deserve to die because of weird biases in the training data that no one is going to notice.

@oakreef definitely both and it will be just as awful as their creators

@amphetamine @oakreef It's not going to be "people in green cardigans" it's going to be oppressed minorities in the country the machine was trained in.

@oakreef @Dingy_Rat 😂😂 yes omg yes they are so so so far away from the problems they imagine and they have no self-awareness of that

@oakreef I am expecting some really weird singularities in the decision process. Like if someone has a left pinky finder that is just the right length they have a weight of negative infinity and the cars feel that they have the obligation to run over this person.
Or if someone has the correct nose length to hair color ratio they must be protected at all costs so cars will actively run away from the person to keep them out of danger at the expense of anything else.

@oakreef we will know car AI has become sufficiently advanced when it starts running over nazis and terfs

Sign in to participate in the conversation
Mastodon

Follow friends and discover new ones. Publish anything you want: links, pictures, text, video. This server is run by the main developers of the Mastodon project. Everyone is welcome as long as you follow our code of conduct!